In this episode of The Acquirer’s Podcast Tobias chats with David Trainer, Founder and CEO of New Constructs. Companies use adjustments and accounting shenanigans to manage or manipulate their earnings often to disguise the true economic picture of what’s occurring in the company. New Constructs unpack those distortions to deliver the real economic reality of the company. During the interview David provided some great insights into:
- New Constructs – Providing One Version Of The Truth When It Comes To Fundamentals
- Most People In Wall Street Think Fundamentals Don’t Matter
- Earnings For S&P 500 Companies Are Distorted By 22% On Average
- There Is A Significant Bias In Terms Of Hidden Items Around Reported Earnings Periods
- Neutron Jack – Find Me Some Unusual Merger, Gains, Expenses Or Restructuring Reserve, Gains, Losses That I Can Use
- Using Machine Learning To Analyse Footnotes
- Typical Tricks That Managers Use To Game Their Statements
- Warren Buffett – Financial Weapons Of Mass Destruction
- Accounting Changes Are Moving The Needle In The Wrong Direction
- Core Earnings In The Tech Sector Are Actually Starting To Fall
- Enron Employed More People In Risk Management Than Any Other Department
- The Whisper Number
References in this podcast:
You can find out more about Tobias’ podcast here – The Acquirers Podcast. You can also listen to the podcast on your favorite podcast platforms here:
Tobias Carlisle: All right. Say when you’re ready.
David Trainer: I’m ready.
Tobias Carlisle: Hi, I’m Tobias Carlisle. This is the Acquirers Podcast. My special guest today is David Trainer, the founder and CEO of New Constructs. Companies use adjustments and accounting shenanigans to manage or manipulate their earnings often to disguise the true economic picture of what’s occurring in the company. New Constructs unpack those distortions to deliver the real economic reality of the company. We’re going to talk to David about how New Constructs does that right after this.
Speaker 3: Tobias Carlisle is the founder and principal of Acquirers Fund. For regulatory reasons, he will not discuss any of the Acquirers Funds on this podcast. All opinions expressed by podcast participants are solely their own and do not reflect the opinions of Acquirers Funds or affiliates. For more information, visit acquirersfunds.com.
Tobias Carlisle: How are you David?
David Trainer: I’m doing great. Nice to meet you. Nice to see you.
Tobias Carlisle: Likewise. Just explain to us what New Constructs is.
David Trainer: Sure. We’ve been around for a long time. I started this business out of my apartment in New York in 2002 and the goal has been to provide investors with sort of the ideal model and scrubbing of financials that they would have if they had an infinite amount of time to read through Ks and Qs and build sort of a supermodel every time. That was my job at Credit Suisse in the mid 90s. I was in charge of a sort of a global template that we integrated and implemented around the world with a small team.
David Trainer: But I saw that there were sort of real limitations to Excel, limitations to just time in terms of how long it takes to read a filing. This was back, when we first started doing this… We first started, when I first started doing this kind of work, Toby, the annual reports were like maybe 20 or 30 pages long and now they are 200 to 2000. I was doing this at Credit Suisse. It was tough work. And then lo and behold, the tech bubble comes along and guess what, Frank Quattrone and his people are not reading 10-Ks and 10-Qs, right? You don’t need to know what’s in the 10-K to sell e-greetings.
Tobias Carlisle: You don’t want to know what’s in them.
David Trainer: Correct. You don’t want people to be looking at those. And so, I can tell a lot of funny stories about how that went, the intersection of what I was doing with those guys. But there was this sort of a lot of apathy around understanding fundamentals. And so, it was clear to me that there really weren’t a lot of people around willing to do this kind of work. And so, really New Constructs was born out of this idea that I had right around 2001, 2002, that we need to get machines to do this kind of work. Because number one, humans don’t want to do it, and number two, it’s very difficult to get any number of humans, I mean small number, large number, to interpret filings and data in a consistent way and a right way.
David Trainer: I’d seen, having built thousands of these models and done this work around the world, I’d seen that there really is one version of the truth and it takes a multi-sector view because when you get caught up too much in the accounting idiosyncrasies between companies in a particular sector versus another sector, you sort of lose sight of what I like to call the immutable underlying economics. And the underlying economics that are consistent across all companies in all sectors are this notion of: a, there’s a certain amount of cashflow relative to a certain amount of capital that’s going into a business.
David Trainer: And whether you’re in the business of insurance or selling widgets or providing consulting services, that concept applies. And so, it’s through that lens that we view all accounting data and you break it down that simply, well, the job, you really can get to one version of the truth. It just takes a lot of really focused brain power which we augment with technology. That’s a really long answer. I’m going to stop there.
Tobias Carlisle: Well, I want to come back to the process and how you use the machine learning. But I just wanted to discuss. You did some research, you partnered with Harvard, MIT to do some research to quantify the impact of the earnings, manipulations or the distortions. It probably doesn’t rise to manipulations, it’s just the management of earnings. And so, what were the key findings of that research?
David Trainer: Well, first of all, I do need to be clear that we did not partner. I know this is details, but my colleagues there at Harvard Business School and MIT Sloan are very particular about the fact that it was a completely and totally independent exercise. They came to me with a problem. People didn’t want to take their class anymore with respect to understanding fundamentals and footnotes. And I said, “Well why?” He said, “Well, because they don’t think that fundamentals matter.” And I said, “You know what? They’re right. Most people in Wall Street don’t think fundamentals matter.”
Tobias Carlisle: Is that a cyclical thing or is that a trend over time? Is that a market high kind of thing?
David Trainer: I wish I had the answer to that question.
Tobias Carlisle: What have you seen over the last 20 years or so?
David Trainer: The cycle has moved in one direction. There’s been no cycle. People care less and less. I remember talking to my old boss Mauboussin after the tech bubble when I was starting New Constructs. Kind of making sure that, “Hey, we’re not going to be going straight back into a world like the one we’d experienced together in the tech bubble where nobody cared about footnote.” There was going to be a return to analytical rigor, which we sort of both were thinking. And that’s never happened.
David Trainer: And I think that we can go off on that for a while, but no, it’s not happened. And that’s why students at Harvard Business School didn’t want to take the business analysis and valuation class. And I said, “Well, the students are right because it takes too much time and it’s too much work for people to do this. But what if I told you there was a technology that gave them the really good data that takes so much time to get for free, for the same price you get the bad stuff.” And he said, “Well, that’s interesting.” And so, yes, they thought the technology was interesting and the logical question that comes out of, hey, there’s this new technology to do fundamental research for you is, well, does it produce data that’s enough or better to improve stock-picking?
Tobias Carlisle: Right. I’ve got a few of the statistics here. You said that there’s been an increase over the last 20 years because that was the period of the study. There are 34% more adjustments over that period of time. And you say that, or sorry, the study says that only 55 cents out of every dollar of adjustment is actually reflected in analyst’s valuations.
David Trainer: Yeah. I mean, that’s why the paper is saying that it’s an increasingly material issue and that markets are inefficient because this data is not taken into account because too few people read footnotes, sort of it’s not a surprise, right? To you and me, really I don’t know many people who read footnotes. Our competitors, typically their data collection process is to employ thousands of people in a third world country. And I always like to say, English is hard enough as a second language, try footnotes English, right?
David Trainer: So, yeah. It’s a big deal. And so, the case study that they wrote about the technology gave birth really to the idea of testing the data. And so they went through and they tested the data independently. And for the academics, this is a big deal because they’re effectively introducing to their colleagues a new paradigm for fundamental data, right? All the research on investing and stock-picking over the last 50 years has been based on Compustat. And what these guys are providing their colleagues is evidence that, well, Compustat’s not as complete or accurate a dataset as you think.
Tobias Carlisle: Is it that the Compustat is accurately reporting what is being reported by the companies themselves? It’s just that that is not a reflection of the economic reality of the companies because they’re able to gain on a quarterly basis.
David Trainer: It depends on what you expect. I mean, Compustat is in the business of collecting financial data. You expect them to collect all the data that they should. What the study finds is that not only is Compustat not collecting data from the footnotes that they ought to, but they are miscategorizing or missing about one in every four or five items on the income statement.
Tobias Carlisle: I see.
David Trainer: I mean, for me the takeaway, Toby, is that what we have been unwilling to recognize is a level of sophistication required to analyze financial statements; and scaling that expertise has been sort of an impossibly big hurdle for people up until this point. And I think this paper proves that New Constructs has successfully met that challenge.
Tobias Carlisle: Well, one of the things that I think that the upshot of what they said was that if you’re long and short, so short, the most manipulated, long, the least manipulated or the ones that had sort of the least adjustments, that was 54 bits of outperformance per month, which is six and a half percent over the course of a year, which is very material. If you compare the market, that’s a big margin to help you get there. The last thing that they said that I thought was particularly interesting is that the earnings for the S&P 500 were distorted by 22% on average last year, which is material.
David Trainer: Yeah, I mean, this is a serious deal. I mean, the other thing that the papers shows that there is… the statistics show that there is significant bias in terms of the number of hidden items in core earnings around near beats and meats of earnings. So effectively the papers showing that managers intentionally bury stuff so that they can manipulate earnings, and it’s a large amount. I mean, if Compustat’s missing 45 cents on every dollar, 22% is kind of a… Some stuff ends up netting out Toby. That’s why that number is a little bit lower.
Tobias Carlisle: I see. The last thing in the paper that I thought was interesting was that it’s the most amount of adjustments that we’ve seen since 2000. That’s why I asked you before if this is a cyclical thing, if this is something you see sort of closer to market peaks.
David Trainer: You’re right. And I misunderstood your question. I thought the question was, is it a cyclical thing that people don’t pay attention to footnotes?
Tobias Carlisle: I’m guessing that’s going in one direction.
David Trainer: Correct. Correct. With our technology, they never have to. The level of distortion is cyclical we have found. Looking at the S&P and the overall market in general, the level of distortion that we are seeing as of the third quarter of 2019 is that it is, yeah, it’s not been this high since right before the financial market crisis and the tech bubble.
Tobias Carlisle: That’s kind of an interesting… I just wanted to ask you about, this is sort of a bit of a non sequitur but when you were talking about earnings adjustments, I thought of Neutron Jack and General Electric. Did you ever spend any time with those? Because they were famous for almost beating or just beating on meeting expectations every quarter. Is that something that you ever looked at?
David Trainer: Yeah. In fact, we wrote a long report on GE about a year or two. Let me look it up exactly when we wrote this report predicting that GE was going to see a potentially very large… we wrote an April 6, 2016. We said that we think that the implied share price of GE was closer to $18 a share because of how overvalued the company looked relative to its return on invested capital. And when GE ended up crashing, I guess a year, a year and a half or so ago, it got right back around to about that level.
Tobias Carlisle: Right.
David Trainer: And so, what we found is that yeah, too much of the street was buying in to the proforma core earnings number when you took a look at the underlying economics that the stock was significantly overvalued. I have people ask me all the time when we bring up these kinds of situations, “Okay, yeah, I see it, but when’s it going to crack? When will we see the reconciliation?” And sometimes these disconnects persist for years as they did with GE.
David Trainer: It was always funny to me because when I was on Wall Street too, everyone loved Jack Welch. He was the man from the gut. He was a good guy. I saw him admit on CNBC one time. He said it like this. He goes, “You know.” They said, “What’s your secret to doing so well with earnings all the time.” He said, “You know what? I’ve got this little division called M&A. And whenever I need a penny here or a penny there, I just go to those guys and say, find me some unusual merger or gains or expenses or restructuring reserve, gains, losses that I can use.”
David Trainer: And I thought to myself, if that’s not a dead giveaway, like this is a game, I don’t know what is. But yeah, GE is a great example of where for so long the mentality of the street had so much bought into sort of this great GE thing when really the emperor had no clothes and it took a long for the market to figure-
Tobias Carlisle: It took a long time.
David Trainer: Right.
Tobias Carlisle: So what’s the process at New Constructs? You have this machine learning process that goes through the notes and then there’s a human being looks at it. Is that how it happens?
David Trainer: That’s roughly it, right. And it’s built on… The most important thing to think about here is how long it took us to get to that point. I mentioned I started this business out of my apartment in 2002 and I spent about a year figuring out whether or not the technology to do what I wanted to do was possible. The original technology was really just an application that combined the filing itself; because traditionally you got to download a file and you look at the numbers and you type them into Excel. And every time you finish typing one number into Excel, all the intelligence that went into figuring out where you got it, what it means and why you put it in that cell is gone.
Tobias Carlisle: Right.
David Trainer: And that was a real problem with me when I was at Credit Suisse because occasionally we make mistakes or counting methodologies would change or we need to change data. And the whole idea of how hard it would be to change lots of models at one time usually meant, Toby, that you argue that it didn’t need to be changed because doing that kind of work was crazy. I mean, imagine if you got a thousand of these models and as I did at the time, I had stacks of filings, paper filings because they weren’t in digital form, it’s all 98 and I had been doing this stuff starting in 96. Stacks of filings almost as tall as me.
David Trainer: I got to go through that. I got to find the right filing. I got to look up the area in that filing where this data point exists and I go back and reconcile that to what I put into an Excel model. It takes more time than actually building the model to begin with. So our first innovation was to combine the filing and the database and the parsing tool all in one application so that every time one of our expert analysts parsed a data point from a filing into a bucket of our system, for example, accounts receivable from the balance sheet goes into the accounts receivable bucket in our system.
David Trainer: The machine was not only sort of tracking what the human does but recording the original text, the value, the location and the filing, right? We had to do a lot of that in the beginning before we could go to the machine and say, “Okay, well, a human has parsed accounts receivable from the balance sheet into the same bucket 999 times out of a thousand times that we’ve seen it. Do you still need the human to do this for you, Mr. Machine?” The machine says no. We say, “Okay, great.” And then we offload gradually more to the machine and spend more time on more sophisticated or challenging complex disclosures as well as cover a lot more companies. The first few years we were around, all we covered was the S&P 500.
Tobias Carlisle: Right. And how far have you extended that now?
David Trainer: We covered the top 3000 stocks in the market. In the beginning we didn’t cover Qs either. So we cover all the Qs and the Ks at the top 3000 in the market. Quite frankly, that’s pretty easy for us, right? Our 10-K filing season used to be this six to eight weeks of hell grind going through all these files and now it takes us less than a couple of weeks and people don’t have to come in at six and leave at eight. It’s gotten a lot faster. Really what we’re focused on now is really continuing to perfect the automated parsing processes so we can start covering international companies. Because we’re focused on the US now.
Tobias Carlisle: What’s are some of the typical tricks that managers like to pull that you see to game their statements?
David Trainer: Well, a lot of it’s unusual gains and losses that are buried and things like cost to goods sold or depreciation and amortization or SGNA. And so, you can have changes in restructuring reserves. You can have a loss on the obsolescence of inventories, sort of one of my favorite ones from a long time ago. There’s sort of an almost, I don’t want to say innumerable, but there are a huge number of these sort of unusual items and we have systematically collected all these over time so that our dictionary of these is really large.
David Trainer: And that’s a big trick here is that so much of machine learning, the experts will tell you, is about the quality and the size of the training dataset. And so, we spent really a lot of the first few years at the company just really building out that training dataset. It’s really good and sophisticated. And as the paper from the Harvard and MIT guys points out, it’s not just that we collected the data, but it’s that we correctly categorized the data.
David Trainer: I think, in my opinion, that’s only possible with an expert because there’s just so much stuff and so much variation. If you don’t have somebody who knows how to build a model collecting data, you’re probably not going to get the data right. And all of our analysts go through six plus months of training before they really are allowed to collect data on their own.
Tobias Carlisle: How many analysts do you have?
David Trainer: We’ve got a team of about eight analysts right now. We do all of this. We’ve been as many as 15 analysts, right? So we’re doing this with less numbers but there’s so much augmentation in the machine too. A big part of the process are sort of left out… I don’t know, maybe it’s only interesting to nerds like me, but we have a ton of these little [crosstalk 00:20:37]-
Tobias Carlisle: … listeners are definitely nerdy enough to enjoy it. So go for it.
David Trainer: Great, great. A big part of what we built after a couple of years was what I call data checks. These are just little things that the machine comes behind the human after the initial parse and says, “Hey, this number that you parsed in 2018 is 10000% bigger than 2017. Do you think maybe you collected it in…”
Tobias Carlisle: You got the decimal place wrong.
David Trainer: Correct. Right. Just little stuff like that. Or, “Hey, the income statement is off by $34, which is exactly the same amount as discharge. Did you collect it as a positive when it should’ve been as a negative?” All kinds of little stuff like that that I don’t want to distract my smart expert analysts with having to think about when they go to a filing. Because guess what? If it’s something that a machine can figure out, let’s impart that to the machines and reserve the intelligence and attention span and focus of our analysts to the things that are interesting and challenging. And that’s a huge part of our culture and our focus because otherwise I don’t think I can keep these folks around long enough to justify how long it takes to train them.
Tobias Carlisle: Would you make any changes to gap accounting? Do you think that… is there anything that you would do differently? I’m taking that that’s a yes.
David Trainer: Oh yeah. I spent five years on FASB’s investor advisory committee and it gave me a real appreciation for the depth of thought and effort that the board puts into accounting rules. The level of complexity and the challenges of changing and dealing with these things across so many companies. And it’s something I sympathize with because that’s a lot of what New Constructs does, right? We’re trying to create a one version of the truth: measure profitability, return on invested capital is what we call that. And what I referred to before as cashflow from the business relative to the capital in the business. It’s really hard thing to do.
David Trainer: And so, yes, the challenge with the accounting experts that you have at FASB is they’re not investors and their history, if you look through it, that group of folks was for a long time sponsored entirely by corporate America. So, there’s some conflicts there. Now, they’re not anymore, but the culture is still evolving. And so, I spent a lot of time really championing the investor perspective because every time that FASB says, “Oh, here’s something we can fix,” and they go to corporate American and corporate America says, “Oh no, we can’t do that. It’s too expensive.”
David Trainer: And I’m like, of course they’re going to say that. They don’t want it to be there. Corporate America sometimes excuses are like, “Yeah, well, we can’t do that. We don’t have the data for that.” And I always would say, “Okay, that’s information too.” In particular, this excuse was around some additional derivative disclosures. I told FASB, I said, “This is a very troubling response because we need this information as investors to understand the relative risk involved in their derivative portfolio.”
David Trainer: So if the company is coming back with, we don’t have that data, it’s too expensive for us to pull that together, then that tells me they don’t know what’s going on in their portfolio, which is, Warren Buffett has called these weapons of mass financial destruction. It seems to be a highly reckless and careless way of managing your business. So, either they do have it and they’re running their business right. In which case there’s no real cost to sharing it. Or they don’t have it and they’re not running their business right. In which case, I need to know that too.
Tobias Carlisle: I know.
David Trainer: So…
Tobias Carlisle: Sorry. Keep going.
David Trainer: Yeah, I’m rambling too much. The takeaway was, I had to have that conversation a few too many times to be honest. And so, yeah, I’ll pause there but I can give you some real specifics on even some recent accounting changes that are moving the needle in the wrong direction.
Tobias Carlisle: Yeah, please.
David Trainer: There’s a recent one with respect to unrealized gains and losses. Warren Buffett has hammered on this one as well. And for a long time FASB and the rule said, “Unrealized gains and losses related to your investment portfolio will not flow through the income statement.” Because as we all know, especially quarter to quarter, asset prices fluctuate and the fluctuation of asset prices should not really affect someone’s interpretation of the underlying economics of the business, especially if you’re Intel or especially if you’re Apple. You’ve got a lot of cash on the balance sheet that’s going to be inevitably moving because you got to put it in, you got to invest it.
David Trainer: That’s the way it used to be, and all this would flow through to the balance sheet and show up in accumulated other comprehensive income, which would fluctuate a lot. And so, we wouldn’t have to make an adjustment for this in our NOPEC calculation, but we would pull out the accumulated OCI from invested capital so we could take the swings out of the denominator. But FASB recently changed the rules and said, “No, no. We want all those unrealized gains and losses to flow through.” That’s great. Bad move.
Tobias Carlisle: Who does that help to do it that way? Because I saw Buffett complain. I was at the AGM, I was at the Berkshire general meeting and it’s something that he’s mentioned a few times as distorting what they do. Who does that help to report them that way?
David Trainer: I always ask that question too, Toby, because that is the question. All right, somebody is benefiting from this, right? Somebody… I’m not really sure who. I mean, maybe I guess people who’ve got gains and want to be able to boost their numbers on that. Could be… look, I’ve been doing this for a long time and I’ve seen a lot of efforts to obfuscate performance over the years. Generally what this rule does is just introduce more noise and generally undermine the quality of accounting data, which creates more room for misinformation, which is a big part of what most of Wall Street’s trading operations are based on.
David Trainer: I look at CNBC and Cramer, these are propaganda misinformation machines based on, there’s a lot of subjectivity and misinformation. That creates dislocations in asset prices that sophisticated trading shops prey on. I mean, that’s their bread and butter. So, the more churn and the more blood you can throw into the water, the more those guys can make money.
Tobias Carlisle: One of the articles that you publish on the site was some of the most distorted companies or the companies with the most distortions. And you compare that with some of the companies with the fewest distortions or the companies that understate their earnings. One of the ones I found particularly interesting was STZ, it’s a ticket constellation brands which is something that, with full disclosure, we’re short in the firm. Are you able to go through how they’re doing or what they’re doing?
David Trainer: Yeah, absolutely. I’ll pull up the model right away. A big part of all this work we’ve done on the collection side, Toby, to ensure that we can audit any number that goes into the model or the database because that’s the only way you can go back and fix stuff, right? I mean, the level of difficulty and time required to go back and look up a number in a paper filing or even a digital filing. We have all this stuff linked up. We share that with our clients. That’s been sort of a hallmark of our offering to our fundamental PM clients is that every single number is auditable.
David Trainer: When I’m looking at my constellation brands model, we’ve got a tab in there that’s a reconciliation tab that goes through and shows every single adjustment we make to report a net income in order to get to net operating profit. And then every single adjustment we make to take the balance sheet and could convert it into invested capital. And so, when I’m looking at constellation brands, and I’ll just focus on the 2019 fiscal year end data, I can see 3.4 billion in net income and then we’ve got 1.9 billion in non-operating gains hidden.
David Trainer: Hidden items in our system is always stuff that you can’t find on the income statement. It’s either in the balance, oh, I’m sorry. Either on the footnotes or the MDNA. We’ve got reported non-operating items or net expenses of 368 million. Those numbers, you can’t find. The other big adjustment we’re making for, excuse me, constellation brands is 132 million in tax adjustment. I found pretty early on in the system, Toby, we had to do a lot of work on taxes because the income tax provision is often entirely irreconcilable to the real cash operating taxes of the business.
David Trainer: Even if you’re looking at the cashflow statement and the net deferred tax liability footnote, a lot of it is because of this magical item called the valuation allowance, which is something that fluctuates a lot around what the auditor’s perspectives are on the likelihood of the company to pay taxes or to realize tax loss carry forwards. Anyway, it’s highly subjective. I’ve never met anyone who could figure it out.
David Trainer: And we spent a lot of time and a lot of our models were broken in the beginning because we just… there’s a lot of methodologies in a lot of the sort of famous books, the McKinsey book on how to calculate cash operating taxes. We found that that just broke down, Toby, it broke down the ability to get to a real cash operating earnings number by adjusting the income tax provision. Just doesn’t work for some companies.
David Trainer: So anyway, those are the adjustments and if we wanted to go through and look in the footnotes or the filings to see where we found these income statement adjustments; I’ve got this little ability to, not little ability, this awesome ability to go through and click on a tab in our filings or in our models that says marked up filings. And when I go there, it shows me the 2000 10-K, 2019 10-K. It has a section for income statement adjustments. And in there I’ve got a section for the hidden items and I can click on 4.9 million. Flow through an inventory step up and that we find on page 32, which is a footnote that shows some of the components of cost of goods sold.
David Trainer: We’ve got that item. We’ve got a loss on inventory write down. We’ve got other losses. We’ve got an impairment of intangible assets, restructuring and other strategic business development costs. These are all buried inside a cost of goods sold as broken out on page 32. We’ve also got an unrealized net gain on security’s measured at fair value of 1.9 billion.
Tobias Carlisle: Yeah, that’s pretty material.
David Trainer: … Page 111. And we do the same thing for all of our balance sheet adjustments.
Tobias Carlisle: So a lot of these things are just accounting arcana but some of these, is it just that some of this is judgment call and some of this is… or is it that they are not necessarily in relation to constellation brands just generally speaking? Is it that they are trying to give a more positive view of the company then is actually the case?
David Trainer: Without being able to ask management directly, it would be hearsay and I can’t always, I don’t know the motivations of companies. I can tell you that having been looking in footnotes and doing this kind of work since 1996, I’ve found that there appears to be, at least in my opinion, a pretty clear attempt to obfuscate more than to disclose. That experience was sort of confirmed with my experience on FASB when these companies make up these excuses as to why they couldn’t disclose things that they needed to run their business. I think we also see it with XBRL, Toby, right? The number of custom tags that companies have developed, an XBRL effectively creates a digital haystack.
Tobias Carlisle: So you don’t think that… That’s kind of an interesting question. Does that XBRL make it… Just explain what that is and then does that make it easier or harder to pass the statements for a machine?
David Trainer: For us, it took a long time, but it did make it easier. It only took a really long time because we really had to figure out how companies game that system too. And we could never have done it if we hadn’t effectively been tagging all this stuff the right way for a long time. And so, we really had to become real experts in XBRL in order to identify how far up sort of the taxonomy chain companies were messing with stuff. And so, most of XBRL, we have to throw out because companies get it wrong.
Tobias Carlisle: Most you have to throw out.
David Trainer: Most of it we throw out because it’s wrong or it’s customized in some way that makes it not consistent with the taxonomy that’s set out.
Tobias Carlisle: Right.
David Trainer: But XBRL to give some quick background, it stands for eXtensible Business Reporting Language and this idea came around from the AICPA, the big accounting association, around the time I started New Constructs. The idea was that, “Hey, we know all this data comes in all these different forms. Let’s create a standard reporting format so people won’t have to go out.”
Tobias Carlisle: It’s a great idea.
David Trainer: Great idea. Great idea. The fundamental flaw in the strategy and the execution was that they were depending upon the companies to cooperate and collaborate. And you do have a few companies collaborate. Microsoft, they’re awesome, right? Because Microsoft wants everyone to know how profitable they are. But otherwise, the reason I mentioned the number of, I think it’s something like 40,000 different tags now in XBRL. The reason I mentioned that is because having that many different tags to describe what’s going on with the company utterly defeats the purpose of XBRL. You put things into a machine readable form so you can trust a machine to read, not one of them, but all of them. Because if you have to go through and check each one, well, you might as well just look at the darn filing, right?
Tobias Carlisle: Right.
David Trainer: And so, the fact that companies have so deliberately done this and oftentimes deliberately I think create custom tags where they don’t need to. And I can say that I think with a little bit more experience than other folks because we have created a normalized system for understanding core earnings that independent researchers at fairly prestigious places, right, have come along and said, “Hey. By the way, these guys are doing all this and they’re doing it right. And by the way, you can make money if you use this superior dataset.” So we’ve done it.
David Trainer: And I think in some ways that really speaks to the challenge and the magnitude of the problem we solved at New Constructs, Toby. Stop me if I’m getting off on a bad track here, but I think that, and I saw this when I was at Credit Suisse, somewhere along the line, somebody at a firm like us or Compustat or Capital IQ or FactSet, somebody collecting data, someone at XBRL. Someone along the line somewhere needs to say, “Hey, this data point gets treated in this way.”
Tobias Carlisle: Right.
David Trainer: And you really need to have someone get that right. Because if it’s not right every time, then what’s the point of the system? And by being a smaller, private firm, where a very small group of people, most of the time me, are making that decision, well, we can’t get this stuff pushed through. But when you’ve got companies who are going to of course disagree on what one letter means versus another, when you’ve got folks in a third world country who aren’t going to really understand the subtleties of the different things, because there are a lot of accounting items, Toby, that are the same exact word, but they mean different things and vice versa.
David Trainer: If you’re a sell side firm where there’s a lot of expertise, there’s a disincentive anyway because let’s face it, it’s going to be a whole lot harder, even harder to sell WeWork or Lyft or Uber to the public if everyone’s able to look at their profitability on an Apple’s to Bapple’s basis as opposed to a community adjusted basis, right?
Tobias Carlisle: Right.
David Trainer: I mean, think about that. So, the expertise to do this kind of work doesn’t exist in a lot of places. It’s different from the accounting expertise that the folks at FASB need, because you really need to understand how an investor uses data and it’s got to be a streamlined decision making process. That can quickly apply to lots of companies and just a lot of firms aren’t set up for that.
Tobias Carlisle: It’s interesting that you raised Microsoft as somebody who’s assisting in that process because one of the articles that you have says that big tech tends to lead the overstatement of accounting. And I think you gave us, there are a few examples in there, Apple and Facebook and MCHP Microchip, I forget the full name of that. But do you see… does big tech tend to be one of the worst abuses? Is that what you found?
David Trainer: You know what’s interesting? Big tech is leading the decline in core earnings these days. But for a long time, big tech was carrying the majority of the load as well. And so, what we’re really seeing is some reversion to the mean here. Apple in particular had an astronomically high return on invested capital, which I famously pointed out was going to decline and the stock price would decline with it. But I was way wrong about that. I was not wrong about the reversion to the mean of the return on invested capital. They were up in the 160% range.
David Trainer: In a world where you’ve got open competition, that’s just not going to last. Turns out it hasn’t really mattered that much because it’s happened slower I think than maybe I expected. But it’s happening now. Facebook, same thing, because the regulators are… even where competition, in the case of the Apple, they’re just seeing more competition, right? Whether it’s Huawei, whether it’s Samsung, and we haven’t really seen any breakthrough innovations in the iPhone in a long time. In some cases they’re just catching up with where Samsung already is. That’s not to diminish the cash flow generating power of Apple. It’s just that the returns instead of being 360% are now closer to 180%.
David Trainer: And that means that the core profitability of the business is declining. It’s still crazy profitable. Same is true with Facebook. And for a long time, those guys were meant… those companies meant that the tech sector was the only sector where we were seeing a consistent rise in earnings because for the last few years we’ve seen economic earnings for the overall market in decline even though accounting earnings have been going up. And this is the first quarter where we have seen core earnings for the tech sector actually fall, and fall by more than any other sector as well.
Tobias Carlisle: But that the gap earnings don’t look that way. The gap earnings are up for the tech sector.
David Trainer: Oh yeah, that’s right.
Tobias Carlisle: It’s an interesting… that crossover I think is very interesting because that possibly shows that some of the innovation is now coming into the accounting rather than the technology.
David Trainer: Yeah. I mean, isn’t that always the way it goes? I mean, I remember that with Enron. For a while they had a good business. When they ran out of a real business, they just turned to accounting. I don’t know if most people know this or not, but it’s one of my favorite facts about Enron, at the end they employed more people in their risk management division than any other part of the business. And risk management stated thesis or stated mission was to basically make accounting earnings look as good as they could so that managers got paid.
Tobias Carlisle: Yeah, it’s fascinating. Just while we’re talking tech, tell some FrankQuattrone.com stories.
David Trainer: Frank was a little bit above my pay grade. He didn’t make it to New York very often, which is where I was.
Tobias Carlisle: He was at Credit Suisse before he set up his own shop, right?
David Trainer: Correct. Correct. And I was at Credit Suisse. I was at Credit Suisse before, during, and after the tech bubble. A couple of funny stories. I was… This product we had at Credit Suisse was called the value dynamics framework and effectively it meant everybody around the world was going to use this one version of the truth model that focused on return on invested capital. And I probably had 70 to 80% of all the analysts around the world on board with this.
David Trainer: I’d been around Europe and Asia and taught people to do it and had analysts doing a lot of input on their own and it was going great. And Mauboussin was running the morning call and we pretty much didn’t let people on the morning call if they didn’t speak through those terms. If they didn’t talk about returns on capital and they didn’t talk about the expectations for future cashflows baked in the stock prices. Didn’t have to be the only thing they talked about, but they did need to check that box. And not everybody complied but most people complied.
David Trainer: Well, we wake up one morning and Brady Dougan has announced that the team from Deutsche Bank, Quattrone’s team from Deutsche Bank, it may have been UBS, I don’t remember which one, but one of those firms. We’d just been acquired by New Constructs, I’m sorry, by Credit Suisse and it was run by Frank Quattrone. The size of the research department doubled overnight. And for a while, we didn’t always let these guys in the morning call because they couldn’t speak the language, right?
David Trainer: And then I think Brady Dougan called Al Jackson, who was the global head of research and said, “Are you kidding me? Get these guys on the call. This is the cash cow.” And they went on to make billions. And I’ll never forget though, Al Jackson was the global head of research and he was my boss at the time. I remember going to the morning call with him every day, not with him, but I would just go every day. And we saw… when we were forced to bring the tech guys on the morning call, we saw the analytical trend go down real fast. We went from return on capital and expectations baked into stock prices, quickly to price to earnings, then the price to sales, then the price to clicks, and then the price to eyeballs.
David Trainer: And I’m thinking to myself, well, and back then we didn’t have Google analytics. I’m like, “How did they measure all these clicks?” And then I’m like, “All right. I know they can’t measure the eyeballs, right?” But I’ll never forget the first time one of the tech analysts went with price to eyeballs. Al Jackson stands up in the middle of the morning call room, we’re like 30, 40 people in there and throws his papers down on the desk and says, “I can’t believe this blankety blank blank blank,” and kind of curses everyone out of the room and walks out. And I thought, bucket list check, I’m a first real Wall Street move experience. That was one of my favorite stories. A couple other, one other story if you’d want me to go was-
Tobias Carlisle: Yeah, sure.
David Trainer: … in meeting with the directors of research for Quattrone’s tech team. Couple of guys and I won’t mention any names here, but they were the directors of research. There were two big tech analysts that covered some of the biggest names in tech. And so, I did with them what I did with most analysts. I would sit down, I’d say, “Hey, I think you should use this model.” In the beginning I got a lot more pushback than I got toward the end with a traditional analyst. And this was my first, and I had done a lot of one-on-one meetings with the tech analysts, individual analysts.
David Trainer: Most of those meetings, I felt like I would win those meetings. It’s very hard to argue against sort of the merits of what we’re doing. “Hey, we want to accurately express the underlying economics of the business and then we want to accurately quantify what those future economics have to be to justify the price, and speak about valuation in those terms because it’s a lot more tangible way to identify what’s overvalued and undervalued. There’s no reason to be a fortune teller. Mr. market is our fortune teller every day.” Anyway, that’s my pitch.
David Trainer: I was having some success. Eventually I got pushed up to the heads of research and explain to them, “Hey, this is what the model does. Here’s what we do.” And they said, “Well, you know Dave? We can’t use your model.” I said, “Okay. Well, why?” And they said, “Well, it’s because we really have two earnings numbers, the one that we publish and our real one.” I said, “Okay, what do you mean?” He goes, “Well, look, we all know that in order for company stock to go up, it’s got to beat the number. So for the stocks we have buy ratings on, our published earnings estimate is actually a lot lower than we think the earnings power is going to be.” I said, “Whoa, okay.”
David Trainer: And this was before anyone knew about the whisper number, right, or that the whisper number was a well known thing. And I said, “Okay. Well, why don’t we just put whatever your published number is in there? It’s no big deal, right?” They said, “No, no, no. It’s a problem because if we were to put into our cashflow forecast, the cashflow is based on the lower number, well, the stock in your reverse DCF model, it’s going to look expensive because it’s going to take a long time for the cashflows to equal the amount that the stock prices is reflecting.” And I was like, “Okay. I don’t know what to say to that.”
David Trainer: You’ve got two sets of books here. I remember going back to Mauboussin and telling him like, “Hey, this is what they just told me.” And Mauboussin’s first response was, “Have you told Al Jackson?” It turns out that that’s really not the most ethical way to sort of run your research business. And I didn’t have an answer for that. Outing them on sort of a different number that they were public with versus a number that they were private with. This is all since been remedied by regulation FD or Fair Disclosure where you kind of, you can’t really do that anymore. But yeah, that’s a pretty good dirt, I guess.
Tobias Carlisle: Quite a few of those guys got in trouble, including Quattrone and numerous others who we probably shouldn’t name. If folks want to get in contact with you, David, what’s the best way to go about doing that?
David Trainer: www.newconstructs.com. That’s our website. There are plenty of places to get samples of data to demo what the website does, get free research. Look, we’re really trying to minimize the gap between where people’s research or analytical habits are today to where they might want to be. We’re here really to democratize access to the truth behind the numbers. We think it makes the market more efficient. We think it improves the integrity of the capital markets. We think that raises standards of living for people here in the United States and worldwide. And we think it’s the right thing to do.
David Trainer: We think that if not, if it weren’t New Constructs creating this technology to read footnotes, someone else would. Some point in time, we’ve got to get machines to do it because just like digging ditches, humans just don’t want to do it. We’re not here to say, Toby, that fundamentals need to be 100% of your process, but we’re just saying they don’t have to be zero, right? Just like the Harvard Physical students. You don’t have to ignore the real fundamentals. We’re going to give it to you, take it. And it doesn’t have to be everything you do.
David Trainer: In fact, one of our first big partnerships was with Scottrade and I was very candid with those guys on the beginning like, “Hey, maybe this doesn’t work for you because most of your clients focus on technicals.” And they said, “No, no. We want to try to get people more information.” I said, “Well, that makes sense because if you’ve got a list of 10 great technical ideas, why not screen that against fundamentals so you’ve got five great ideas with good technicals and fundamentals. Have your cake and eat it too.” So, we do our best to try to make that as easy as possible for people.
Tobias Carlisle: Yeah, I’m a believer. I’ve spent a lot of time doing it by hand so I certainly I’m a subscriber to your service and I certainly appreciate the numbers because I’m a believer in garbage in, garbage out. So the better the numbers are as they go into the model, the better the model should be.
David Trainer: I appreciate you saying that, and you’re unique in that. Toby, I can’t tell you how many managers I’ve met over the last 15 years who manage 50, $100+ billion who’ve said to me, and this is quoted in the Harvard Business School case study. “Yeah, David, your data’s probably better than what I’m using now, but as long as everyone else has the same bad data, I’m okay with that.” Most of that I got before the ETFs came along.
David Trainer: And so, part of the reason this paper from Harvard Business School and MIT Sloan is so important is that it’s kind of pulling the sheet off the elephant in the room and saying, “Look everybody, there’s no reason to hide now. We know the data’s better and it’s out there and it’s independently. It’s not just David Trainer New Construct saying it. It’s Harvard Business School and MIT Sloan. So, take advantage of it. Why not?” WisdomTree’s a new partner of ours, right? They’re using it in their ETFs. And we’ve got, the business is… this paper has been very good for our business and I think it’s hopefully going to be very good for the markets in general.
Tobias Carlisle: Well, that’s great to hear. And on that note David Trainer, New Constructs, thank you very much.
David Trainer: Thank you Toby. I had a great time.
For all the latest news and podcasts, join our free newsletter here.
Don’t forget to check out our FREE Large Cap 1000 – Stock Screener, here at The Acquirer’s Multiple: